Courtesy of Ylenia Giarratano, Dr Tom MacGillivray and Dr Miguel Bernabeu, this image shows an original optical coherence tomography angiography (OCTA) image of the retina (left panel) acquired with the RTVue-XR Avanti OCT system, and the performance of three automated image segmentation methods used to perform vessel enhancement and binarization (right panels): an optimally oriented flux (OOF) handcrafted filter and two deep learning architectures, U-Net and CS-Net. Evaluation metrics applied to automated segmentation results for retinal scan subimages from 11 individuals found the best performance was achieved by U-Net and CS-Net architectures, and identified OOF as the best handcrafted filter for applications where manually segmented data are not available to retrain those approaches. The source code and the image dataset with associated ground truth manual segmentations have been made openly available to support standardization efforts in OCTA image segmentation.


The image is taken from a recent study published in Translational Vision Science & Technology:

Giarratano Y, Bianchi E, Gray C, Morris A, MacGillivray T, Dhillon B, Bernabeu MO. Automated Segmentation of Optical Coherence Tomography Angiography Images: Benchmark Data and Clinically Relevant Metrics. Transl Vis Sci Technol 2020; 9(13):5.